Evidence-based practice and data interpretation for MRCGP AKT

A trial of a new blood pressure tablet reports that it lowers the 5-year risk of stroke from 8% to 6% compared with placebo. How many people need to take it for 5 years to prevent one stroke?

1. 100 Wrong. You would get 100 from an absolute risk reduction of 1%. Check the subtraction.
2. 25 Wrong. 25% is the relative risk reduction (2 divided by 8). It is not the number needed to treat.
3. 4 Wrong. You get 4 if you use the relative risk reduction (25%) instead of the absolute one. This is the most common mistake.
4. 12.5 Wrong. That is 1 divided by 0.08, the starting risk. The NNT uses the difference between the two risks.
5. 50 correct Correct. The absolute risk reduction is 8% minus 6%, which is 2%. NNT is 1 divided by 0.02, which is 50.

The point: NNT = 1 divided by the absolute risk reduction. Work out the absolute difference first (8% minus 6% = 2%). Relative figures make small benefits look big, so use absolute numbers when you talk to patients.

Source: RCGP curriculum: Evidence in practice, research, teaching and lifelong learning Royal College of General Practitioners · tier 0, exam blueprint / regulator

A trial in adults with COPD compares a new daily inhaler with placebo for one year. Flare-ups happened in 20 in 100 people on placebo and 15 in 100 people on the new inhaler. How many people need to use the new inhaler for one year to prevent one flare-up?

1. 20 correct Correct. The absolute risk reduction is 20 minus 15, which is 5 in 100 or 0.05. One divided by 0.05 gives 20.
2. 25 Wrong. 25% is the relative risk reduction (5 out of the 20 placebo flare-ups). It does not tell you how many people to treat.
3. 5 Wrong. 5 in 100 is the absolute risk reduction itself. The number needed to treat is 1 divided by that figure.
4. 4 Wrong. 4 is 1 divided by the relative risk reduction of 25%. The number needed to treat uses the absolute difference, not the relative one.
5. 35 Wrong. 35 comes from adding the two flare-up rates together, which has no meaning here.

The point: Number needed to treat = 1 divided by the absolute risk reduction. Put the risks in the same form first: 5 in 100 is 0.05, and 1 divided by 0.05 is 20. The relative risk reduction (25% here) sounds bigger and hides how few people benefit. NICE tells clinicians to talk about absolute risk, not relative risk.

Source: NICE NG197: Shared decision making NICE · tier 1, national regulator or guidance

A trial of a new blood thinner for atrial fibrillation reports a major bleed in 3 in 100 people on the new drug over one year. On warfarin the rate was 1 in 100. How many people need to take the new drug for one year to cause one extra major bleed compared with warfarin?

1. 2 Wrong. 2 in 100 is the extra bleeding risk (the absolute risk increase). The number needed to harm is 1 divided by that risk.
2. 3 Wrong. 3 in 100 is the bleed rate on the new drug alone. The comparison with warfarin is missing.
3. 100 Wrong. 100 is the number you would get from warfarin's rate alone (1 in 100). It ignores the extra bleeds from the new drug.
4. 33 Wrong. 33 is 100 divided by 3. That uses only the new drug's bleed rate and ignores the bleeds that warfarin also causes.
5. 50 correct Correct. The absolute risk increase is 3 minus 1, which is 2 in 100 or 0.02. One divided by 0.02 gives 50.

The point: Number needed to harm = 1 divided by the absolute risk increase. Subtract the two event rates first, then invert. It is the same sum as the number needed to treat, but for an unwanted event. Weigh the NNH against the NNT for the benefit the drug gives.

Source: NICE NG197: Shared decision making NICE · tier 1, national regulator or guidance